v0.5

Feature release

General DataModule (#253)

This DataModule supports loading datasets from Hugging Face Datasets format (e.g., "ylecun/mnist") or from local parquet files. It also allows specifying which columns to load and how to transform them using Column objects. It also handles splitting the dataset into training, validation, and test sets based on specified proportions.

Added Visual Transformer (#253)

It takes an image input and produces a fixed-size embedding vector. The CLS token from the ViT model is used as the output representation, and an optional linear projection can be applied on top of it to match a desired output dimension.

Other Changes

• Add validation loss (#252)
• Add GPU in-memory option for DataModule (#246)
• Update model training section in ReadTheDocs
• Clean-up experiment YAML files

Full Changelog: v0.4...v0.5.1

v0.4

Feature release

Separate Spherinator and HiPSter (#230)

To prepare for the modularization of the machine-learning workflow, the HiPSter module has been transferred to a new repository (https://github.com/HITS-AIN/HiPSter). This change enables the use of various training methods and reduces dependencies.

Generic model structure (#229)

The autoencoder architecture was updated to support various data modalities, such as images and spectra.

Other Changes

• Replace poetry, black, and flake8 by uv and ruff (#245)
• Parquet as a general dataset format (#225)

Full Changelog: v0.3...v0.4

v0.3

Feature release

Add user documentation with ReadTheDocs (#140)

Please find at https://spherinator.readthedocs.io/en/latest/index.html

Use Poetry for packaging (#139)

The Python code was restructured and placed in the src directory, with major namespaces for spherinator and hipster included. Poetry can be used to build Python packages and upload them to PyPI. The package installation will include the executables spherinator and hipster making them available in the system path.

Other Changes

• Remove the first version of the hyperspherical distribution (#127)
• ONNX export monitoring (#108)
• Encoder and decoder as model parameters (#98)
• Define a list of IDs for reconstruction callback (#78)
• Replace git submodules with poetry git dependencies (#63)

Full Changelog: v0.2...v0.3

v0.2

Bugfix release and refactoring

See #59 for more details.

v0.1

First prototype presented at ADASS XXXIII